Experimenting with AI in Product Management, but how?

AI is currently the loudest voice in the room. It promises speed, intelligence, and the occasional miracle. For Product Owners and Product Managers, the real challenge is not deciding if AI matters, but figuring out where it genuinely helps instead of quietly adding complexity.

This applies both to the software product itself (How can we integrate AI in a useful way?), but also to the daily work of software product teams (How can we benefit from AI in our daily work?).

The fastest path to clarity is experimentation. Not grand strategies or slide decks, but small, intentional experiments that turn abstract potential into concrete learning: This is a hype. But THAT saves us hours every week.

AI as a technology is just too radical to know how it affects products or product work in advance.

Below are four experiment formats that work particularly well in software teams.

Run a Hackathon and Build AI Prototypes

Hackathons create a rare and wonderful state of mind where people stop worrying about long-term consequences and start exploring. This makes them ideal for AI experiments.

Set aside one or two days and invite teams to build prototypes that use AI either inside your existing product or as a standalone idea. The goal is not production readiness but exploration. Let developers and product people try things that would normally never survive a backlog refinement.

What usually emerges is a much clearer understanding of what is surprisingly easy, what is painfully hard, and what nobody had thought of before. Often the most valuable outcome is not the prototype itself, but the realization that a small AI-powered improvement can outperform a big, ambitious feature idea.

Treat the results as learning artifacts, not commitments. If something looks promising, great. If it fails spectacularly, that is also fine.

Regularly Try One AI Tool in Your Daily Work

Instead of experimenting everywhere at once, focus on one AI tool at a time and intentionally integrate it into your daily routine.

For a few weeks, let AI help with things like writing user stories, summarizing interviews, refining backlog items, or analyzing user analytics. Use it consistently, even when it feels a bit awkward at first. That friction is part of the experiment.

Over time, patterns emerge. You start to see where AI saves real time, where it improves quality, and where it adds more review effort than value.

You will also gain empathy for your users. If an AI tool feels confusing or unreliable to you, it will feel the same to them.

Explore AI Ideas Together With Your Customers

One of the most effective ways to discover meaningful AI use cases is to explore them together with your customers.

Run workshops or brainstorming sessions focused on how AI could support their workflows. Keep the conversation grounded. Instead of asking whether they want AI, ask about repetitive tasks, slow processes, and decisions that require scanning large amounts of information.

Customers rarely ask for "an AI feature". They describe frustrations, shortcuts, and wishful thinking about how things should work. That is exactly where AI can help, if used carefully.

These sessions are also helpful to uncover unspoken concerns around trust, transparency, and control. Addressing those early is far easier than explaining them after a feature is already shipped.

Define a Bold AI Challenge and Commit to It

Sometimes the best insights come from slightly unreasonable experiments.

Define a clear challenge, such as letting AI handle as much product work as possible for one week, or requiring that AI always produces the first version of certain artifacts while humans only review and adjust.

These challenges should be timeboxed and safe to fail. The point is not to prove that AI can replace people, but to discover where it can meaningfully support them.

You will quickly see which tasks AI can fully automate, which ones it can significantly accelerate, and which ones absolutely require human intuition. Expect moments of delight, moments of frustration, and at least one output that sounds confident while being completely wrong.

All of that is valuable learning.

Treat AI Like a New Team Member

AI is not a magic feature you bolt onto a roadmap. It behaves more like a new team member who is incredibly fast, never tired, and occasionally very wrong with great confidence.

By running small, focused experiments, Product Owners and Product Managers can move beyond hype and develop a realistic understanding of where AI creates value in their product and their work.

And the next time someone asks, "So, what's our AI strategy?", you'll have a much better answer than dropping a few buzzwords.